requiring python-weka-wrapper3 >= 0.3.0 now (jpype-based)

using scikit-learn instead of deprecated sklearn dependency for scikit-learn (https://github.com/fracpete/sklearn-weka-plugin/pull/10)

WekaEstimator (module sklweka.classifiers) now has a custom score method that distinguishes between classification and regression to return the correct score.

renamed data to X and targets to y, since some sklearn schemes use named arguments

added dummy argument sample_weight=None to fit, score and fit_predict methods

fixed: when supplying Classifier or JBObject instead of classname/options, classname/options now get determined automatically

method to_instance (module: sklweka.dataset) now performs correct missing value check

method to_nominal_labels (module: sklweka.dataset) generates nicer labels now

WekaEstimator (module sklweka.classifiers) and WekaCluster (module sklweka.clusters) now allow specifying how many labels a particular nominal attribute or class attribute has (to avoid error message like Cannot handle unary class attribute! if there is only one label present in a particular split)

the to_nominal_attributes method in the sklearn.dataset module requires now the indices parameter (incorrectly declared as optional); can parse a range string now as well

fixed the fit, set_params and __str__ methods fo the MakeNominal transformer (module sklweka.preprocessing)

WekaEstimator (module sklweka.classifiers) and WekaCluster (module sklweka.clusters) now allow specifying which attributes to turn into nominal ones, which avoids having to manually convert the data (either as list with 0-based indices or range string with 1-based indices)

set_params methods now ignore empty dictionaries

fixed sorting of labels in to_instances method in module sklweka.dataset

redoing X when no class present in load_arff method (module sklweka.dataset)

added load_dataset method in module sklweka.dataset that uses Weka to load the data before converting it into sklearn data structures (slower, but more flexible)

added support for generating data via Weka’s data generators

requiring python-weka-wrapper3 version 0.2.1 at least in order to offer pickle support

initial release